Research On Nolinear Camera Calibration Based On Feature Point Matching

Vision is a main mean for human to observe and understand the world.Along with the mature of signal processing theory and the popularization ofcomputer application, which leads to the development of computer vision andpromotes the development of machine vision. Camera calibration is the firstimportant step in computer vision research, so the research on camera calibrationtechnology has important theoretical significance and it can improve theprecision of computer vision application and accuracy in the practical application.First of all, the paper discusses the basis of camera calibration model indetail. Based on the Harris/Plessey, SUSAN, MlC corner detection operatortheory, it can be given an improved method that is on the basis of Harris cornerdetection operator. Through the analysis of this method, the sub-pixel levelaccuracy can be obtained.Secondly, based on the analysis of the existing camera calibration model,combining the relevant calibration library function OpenCV open source visuallibrary, proposes an improved camera calibration method. Improved cameracalibration method is accomplished through three steps; First of all, introducestandard homography matrix (the original should be the lower right corner ofnormalized matrix elements) in the linear camera model, using the least squaremethod for solving statically indeterminate equations, to calibrate the cameraintrinsic matrix and the initial value of the external parameter matrix; Secondly,the nonlinear model is introduced, in nonlinear models mainly consider two kindsof distortion model (tangential and radial), then the intersection point coordinatesof image distortion and the first step for the camera intrinsic matrix and theinitialparameter matrix, initial variable parameters by least square method to solve thedistortion; Finally, nonlinear optimization of all parameters, optimization of allinternal and external parameters of the camera eventually need. Finally, with the utilization of VC++a camera calibration program can bewritten for invoking the OpenCV open source visual library. Through thisprogram it can solve the calibration image acquisition, image corner extractionand accurate camera calibration parameters quickly and effectively.This paper is mainly about the basic theory knowledge and related methodof camera calibration. We build the corresponding relation between the3D pointsand2D image points through imaging model and solve the correspondingrelationship to get the camera calibration parameters. The purpose of this paper isproviding a reliable theoretical support and data support for follow-up study ofcomputer vision and machine vision.